projects/linux·blob master · refs

838 lines · 21.3 KiB

1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/mm/mlock.c
4 *
5 * (C) Copyright 1995 Linus Torvalds
6 * (C) Copyright 2002 Christoph Hellwig
7 */
8
9#include <linux/capability.h>
10#include <linux/mman.h>
11#include <linux/mm.h>
12#include <linux/sched/user.h>
13#include <linux/swap.h>
14#include <linux/swapops.h>
15#include <linux/pagemap.h>
16#include <linux/folio_batch.h>
17#include <linux/pagewalk.h>
18#include <linux/mempolicy.h>
19#include <linux/syscalls.h>
20#include <linux/sched.h>
21#include <linux/export.h>
22#include <linux/rmap.h>
23#include <linux/mmzone.h>
24#include <linux/hugetlb.h>
25#include <linux/memcontrol.h>
26#include <linux/mm_inline.h>
27#include <linux/secretmem.h>
28
29#include "internal.h"
30
31struct mlock_fbatch {
32 local_lock_t lock;
33 struct folio_batch fbatch;
34};
35
36static DEFINE_PER_CPU(struct mlock_fbatch, mlock_fbatch) = {
37 .lock = INIT_LOCAL_LOCK(lock),
38};
39
40bool can_do_mlock(void)
41{
42 if (rlimit(RLIMIT_MEMLOCK) != 0)
43 return true;
44 if (capable(CAP_IPC_LOCK))
45 return true;
46 return false;
47}
48EXPORT_SYMBOL(can_do_mlock);
49
50/*
51 * Mlocked folios are marked with the PG_mlocked flag for efficient testing
52 * in vmscan and, possibly, the fault path; and to support semi-accurate
53 * statistics.
54 *
55 * An mlocked folio [folio_test_mlocked(folio)] is unevictable. As such, it
56 * will be ostensibly placed on the LRU "unevictable" list (actually no such
57 * list exists), rather than the [in]active lists. PG_unevictable is set to
58 * indicate the unevictable state.
59 */
60
61static struct lruvec *__mlock_folio(struct folio *folio, struct lruvec *lruvec)
62{
63 /* There is nothing more we can do while it's off LRU */
64 if (!folio_test_clear_lru(folio))
65 return lruvec;
66
67 lruvec = folio_lruvec_relock_irq(folio, lruvec);
68
69 if (unlikely(folio_evictable(folio))) {
70 /*
71 * This is a little surprising, but quite possible: PG_mlocked
72 * must have got cleared already by another CPU. Could this
73 * folio be unevictable? I'm not sure, but move it now if so.
74 */
75 if (folio_test_unevictable(folio)) {
76 lruvec_del_folio(lruvec, folio);
77 folio_clear_unevictable(folio);
78 lruvec_add_folio(lruvec, folio);
79
80 __count_vm_events(UNEVICTABLE_PGRESCUED,
81 folio_nr_pages(folio));
82 }
83 goto out;
84 }
85
86 if (folio_test_unevictable(folio)) {
87 if (folio_test_mlocked(folio))
88 folio->mlock_count++;
89 goto out;
90 }
91
92 lruvec_del_folio(lruvec, folio);
93 folio_clear_active(folio);
94 folio_set_unevictable(folio);
95 folio->mlock_count = !!folio_test_mlocked(folio);
96 lruvec_add_folio(lruvec, folio);
97 __count_vm_events(UNEVICTABLE_PGCULLED, folio_nr_pages(folio));
98out:
99 folio_set_lru(folio);
100 return lruvec;
101}
102
103static struct lruvec *__mlock_new_folio(struct folio *folio, struct lruvec *lruvec)
104{
105 VM_BUG_ON_FOLIO(folio_test_lru(folio), folio);
106
107 lruvec = folio_lruvec_relock_irq(folio, lruvec);
108
109 /* As above, this is a little surprising, but possible */
110 if (unlikely(folio_evictable(folio)))
111 goto out;
112
113 folio_set_unevictable(folio);
114 folio->mlock_count = !!folio_test_mlocked(folio);
115 __count_vm_events(UNEVICTABLE_PGCULLED, folio_nr_pages(folio));
116out:
117 lruvec_add_folio(lruvec, folio);
118 folio_set_lru(folio);
119 return lruvec;
120}
121
122static struct lruvec *__munlock_folio(struct folio *folio, struct lruvec *lruvec)
123{
124 int nr_pages = folio_nr_pages(folio);
125 bool isolated = false;
126
127 if (!folio_test_clear_lru(folio))
128 goto munlock;
129
130 isolated = true;
131 lruvec = folio_lruvec_relock_irq(folio, lruvec);
132
133 if (folio_test_unevictable(folio)) {
134 /* Then mlock_count is maintained, but might undercount */
135 if (folio->mlock_count)
136 folio->mlock_count--;
137 if (folio->mlock_count)
138 goto out;
139 }
140 /* else assume that was the last mlock: reclaim will fix it if not */
141
142munlock:
143 if (folio_test_clear_mlocked(folio)) {
144 zone_stat_mod_folio(folio, NR_MLOCK, -nr_pages);
145 if (isolated || !folio_test_unevictable(folio))
146 __count_vm_events(UNEVICTABLE_PGMUNLOCKED, nr_pages);
147 else
148 __count_vm_events(UNEVICTABLE_PGSTRANDED, nr_pages);
149 }
150
151 /* folio_evictable() has to be checked *after* clearing Mlocked */
152 if (isolated && folio_test_unevictable(folio) && folio_evictable(folio)) {
153 lruvec_del_folio(lruvec, folio);
154 folio_clear_unevictable(folio);
155 lruvec_add_folio(lruvec, folio);
156 __count_vm_events(UNEVICTABLE_PGRESCUED, nr_pages);
157 }
158out:
159 if (isolated)
160 folio_set_lru(folio);
161 return lruvec;
162}
163
164/*
165 * Flags held in the low bits of a struct folio pointer on the mlock_fbatch.
166 */
167#define LRU_FOLIO 0x1
168#define NEW_FOLIO 0x2
169static inline struct folio *mlock_lru(struct folio *folio)
170{
171 return (struct folio *)((unsigned long)folio + LRU_FOLIO);
172}
173
174static inline struct folio *mlock_new(struct folio *folio)
175{
176 return (struct folio *)((unsigned long)folio + NEW_FOLIO);
177}
178
179/*
180 * mlock_folio_batch() is derived from folio_batch_move_lru(): perhaps that can
181 * make use of such folio pointer flags in future, but for now just keep it for
182 * mlock. We could use three separate folio batches instead, but one feels
183 * better (munlocking a full folio batch does not need to drain mlocking folio
184 * batches first).
185 */
186static void mlock_folio_batch(struct folio_batch *fbatch)
187{
188 struct lruvec *lruvec = NULL;
189 unsigned long mlock;
190 struct folio *folio;
191 int i;
192
193 for (i = 0; i < folio_batch_count(fbatch); i++) {
194 folio = fbatch->folios[i];
195 mlock = (unsigned long)folio & (LRU_FOLIO | NEW_FOLIO);
196 folio = (struct folio *)((unsigned long)folio - mlock);
197 fbatch->folios[i] = folio;
198
199 if (mlock & LRU_FOLIO)
200 lruvec = __mlock_folio(folio, lruvec);
201 else if (mlock & NEW_FOLIO)
202 lruvec = __mlock_new_folio(folio, lruvec);
203 else
204 lruvec = __munlock_folio(folio, lruvec);
205 }
206
207 if (lruvec)
208 lruvec_unlock_irq(lruvec);
209 folios_put(fbatch);
210}
211
212void mlock_drain_local(void)
213{
214 struct folio_batch *fbatch;
215
216 local_lock(&mlock_fbatch.lock);
217 fbatch = this_cpu_ptr(&mlock_fbatch.fbatch);
218 if (folio_batch_count(fbatch))
219 mlock_folio_batch(fbatch);
220 local_unlock(&mlock_fbatch.lock);
221}
222
223void mlock_drain_remote(int cpu)
224{
225 struct folio_batch *fbatch;
226
227 WARN_ON_ONCE(cpu_online(cpu));
228 fbatch = &per_cpu(mlock_fbatch.fbatch, cpu);
229 if (folio_batch_count(fbatch))
230 mlock_folio_batch(fbatch);
231}
232
233bool need_mlock_drain(int cpu)
234{
235 return folio_batch_count(&per_cpu(mlock_fbatch.fbatch, cpu));
236}
237
238/**
239 * mlock_folio - mlock a folio already on (or temporarily off) LRU
240 * @folio: folio to be mlocked.
241 */
242void mlock_folio(struct folio *folio)
243{
244 struct folio_batch *fbatch;
245
246 local_lock(&mlock_fbatch.lock);
247 fbatch = this_cpu_ptr(&mlock_fbatch.fbatch);
248
249 if (!folio_test_set_mlocked(folio)) {
250 int nr_pages = folio_nr_pages(folio);
251
252 zone_stat_mod_folio(folio, NR_MLOCK, nr_pages);
253 __count_vm_events(UNEVICTABLE_PGMLOCKED, nr_pages);
254 }
255
256 folio_get(folio);
257 if (!folio_batch_add(fbatch, mlock_lru(folio)) ||
258 !folio_may_be_lru_cached(folio) || lru_cache_disabled())
259 mlock_folio_batch(fbatch);
260 local_unlock(&mlock_fbatch.lock);
261}
262
263/**
264 * mlock_new_folio - mlock a newly allocated folio not yet on LRU
265 * @folio: folio to be mlocked, either normal or a THP head.
266 */
267void mlock_new_folio(struct folio *folio)
268{
269 struct folio_batch *fbatch;
270 int nr_pages = folio_nr_pages(folio);
271
272 local_lock(&mlock_fbatch.lock);
273 fbatch = this_cpu_ptr(&mlock_fbatch.fbatch);
274 folio_set_mlocked(folio);
275
276 zone_stat_mod_folio(folio, NR_MLOCK, nr_pages);
277 __count_vm_events(UNEVICTABLE_PGMLOCKED, nr_pages);
278
279 folio_get(folio);
280 if (!folio_batch_add(fbatch, mlock_new(folio)) ||
281 !folio_may_be_lru_cached(folio) || lru_cache_disabled())
282 mlock_folio_batch(fbatch);
283 local_unlock(&mlock_fbatch.lock);
284}
285
286/**
287 * munlock_folio - munlock a folio
288 * @folio: folio to be munlocked, either normal or a THP head.
289 */
290void munlock_folio(struct folio *folio)
291{
292 struct folio_batch *fbatch;
293
294 local_lock(&mlock_fbatch.lock);
295 fbatch = this_cpu_ptr(&mlock_fbatch.fbatch);
296 /*
297 * folio_test_clear_mlocked(folio) must be left to __munlock_folio(),
298 * which will check whether the folio is multiply mlocked.
299 */
300 folio_get(folio);
301 if (!folio_batch_add(fbatch, folio) ||
302 !folio_may_be_lru_cached(folio) || lru_cache_disabled())
303 mlock_folio_batch(fbatch);
304 local_unlock(&mlock_fbatch.lock);
305}
306
307static inline unsigned int folio_mlock_step(struct folio *folio,
308 pte_t *pte, unsigned long addr, unsigned long end)
309{
310 unsigned int count = (end - addr) >> PAGE_SHIFT;
311 pte_t ptent = ptep_get(pte);
312
313 if (!folio_test_large(folio))
314 return 1;
315
316 return folio_pte_batch(folio, pte, ptent, count);
317}
318
319static inline bool allow_mlock_munlock(struct folio *folio,
320 struct vm_area_struct *vma, unsigned long start,
321 unsigned long end, unsigned int step)
322{
323 /*
324 * For unlock, allow munlock large folio which is partially
325 * mapped to VMA. As it's possible that large folio is
326 * mlocked and VMA is split later.
327 *
328 * During memory pressure, such kind of large folio can
329 * be split. And the pages are not in VM_LOCKed VMA
330 * can be reclaimed.
331 */
332 if (!vma_test(vma, VMA_LOCKED_BIT))
333 return true;
334
335 /* folio_within_range() cannot take KSM, but any small folio is OK */
336 if (!folio_test_large(folio))
337 return true;
338
339 /* folio not in range [start, end), skip mlock */
340 if (!folio_within_range(folio, vma, start, end))
341 return false;
342
343 /* folio is not fully mapped, skip mlock */
344 if (step != folio_nr_pages(folio))
345 return false;
346
347 return true;
348}
349
350static int mlock_pte_range(pmd_t *pmd, unsigned long addr,
351 unsigned long end, struct mm_walk *walk)
352
353{
354 struct vm_area_struct *vma = walk->vma;
355 spinlock_t *ptl;
356 pte_t *start_pte, *pte;
357 pte_t ptent;
358 struct folio *folio;
359 unsigned int step = 1;
360 unsigned long start = addr;
361
362 ptl = pmd_trans_huge_lock(pmd, vma);
363 if (ptl) {
364 if (!pmd_present(*pmd))
365 goto out;
366 if (is_huge_zero_pmd(*pmd))
367 goto out;
368 folio = pmd_folio(*pmd);
369 if (folio_is_zone_device(folio))
370 goto out;
371 if (vma_test(vma, VMA_LOCKED_BIT))
372 mlock_folio(folio);
373 else
374 munlock_folio(folio);
375 goto out;
376 }
377
378 start_pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
379 if (!start_pte) {
380 walk->action = ACTION_AGAIN;
381 return 0;
382 }
383
384 for (pte = start_pte; addr != end; pte++, addr += PAGE_SIZE) {
385 ptent = ptep_get(pte);
386 if (!pte_present(ptent))
387 continue;
388 folio = vm_normal_folio(vma, addr, ptent);
389 if (!folio || folio_is_zone_device(folio))
390 continue;
391
392 step = folio_mlock_step(folio, pte, addr, end);
393 if (!allow_mlock_munlock(folio, vma, start, end, step))
394 goto next_entry;
395
396 if (vma_test(vma, VMA_LOCKED_BIT))
397 mlock_folio(folio);
398 else
399 munlock_folio(folio);
400
401next_entry:
402 pte += step - 1;
403 addr += (step - 1) << PAGE_SHIFT;
404 }
405 pte_unmap(start_pte);
406out:
407 spin_unlock(ptl);
408 cond_resched();
409 return 0;
410}
411
412/*
413 * mlock_vma_pages_range() - mlock any pages already in the range,
414 * or munlock all pages in the range.
415 * @vma - vma containing range to be mlock()ed or munlock()ed
416 * @start - start address in @vma of the range
417 * @end - end of range in @vma
418 * @new_vma_flags - the new set of flags for @vma.
419 *
420 * Called for mlock(), mlock2() and mlockall(), to set @vma VMA_LOCKED_BIT;
421 * called for munlock() and munlockall(), to clear VMA_LOCKED_BIT from @vma.
422 */
423static void mlock_vma_pages_range(struct vm_area_struct *vma,
424 unsigned long start, unsigned long end,
425 vma_flags_t *new_vma_flags)
426{
427 static const struct mm_walk_ops mlock_walk_ops = {
428 .pmd_entry = mlock_pte_range,
429 .walk_lock = PGWALK_WRLOCK_VERIFY,
430 };
431
432 /*
433 * There is a slight chance that concurrent page migration,
434 * or page reclaim finding a page of this now-VMA_LOCKED_BIT vma,
435 * will call mlock_vma_folio() and raise page's mlock_count:
436 * double counting, leaving the page unevictable indefinitely.
437 * Communicate this danger to mlock_vma_folio() with VMA_IO_BIT,
438 * which is a VMA_SPECIAL_FLAGS flag not allowed on VMA_LOCKED_BIT vmas.
439 * mmap_lock is held in write mode here, so this weird
440 * combination should not be visible to other mmap_lock users;
441 * but WRITE_ONCE so rmap walkers must see VMA_IO_BIT if VMA_LOCKED_BIT.
442 */
443 if (vma_flags_test(new_vma_flags, VMA_LOCKED_BIT))
444 vma_flags_set(new_vma_flags, VMA_IO_BIT);
445 vma_start_write(vma);
446 vma_flags_reset_once(vma, new_vma_flags);
447
448 lru_add_drain();
449 walk_page_range_vma(vma, start, end, &mlock_walk_ops, NULL);
450 lru_add_drain();
451
452 if (vma_flags_test(new_vma_flags, VMA_IO_BIT)) {
453 vma_flags_clear(new_vma_flags, VMA_IO_BIT);
454 vma_flags_reset_once(vma, new_vma_flags);
455 }
456}
457
458/*
459 * mlock_fixup - handle mlock[all]/munlock[all] requests.
460 *
461 * Filters out "special" vmas -- VMA_LOCKED_BIT never gets set for these, and
462 * munlock is a no-op. However, for some special vmas, we go ahead and
463 * populate the ptes.
464 *
465 * For vmas that pass the filters, merge/split as appropriate.
466 */
467static int mlock_fixup(struct vma_iterator *vmi, struct vm_area_struct *vma,
468 struct vm_area_struct **prev, unsigned long start,
469 unsigned long end, vma_flags_t *new_vma_flags)
470{
471 const vma_flags_t old_vma_flags = vma->flags;
472 struct mm_struct *mm = vma->vm_mm;
473 int nr_pages;
474 int ret = 0;
475
476 if (vma_flags_same_pair(&old_vma_flags, new_vma_flags) ||
477 vma_is_secretmem(vma) || !vma_supports_mlock(vma)) {
478 /*
479 * Don't set VMA_LOCKED_BIT or VMA_LOCKONFAULT_BIT and don't
480 * count. For secretmem, don't allow the memory to be unlocked.
481 */
482 goto out;
483 }
484
485 vma = vma_modify_flags(vmi, *prev, vma, start, end, new_vma_flags);
486 if (IS_ERR(vma)) {
487 ret = PTR_ERR(vma);
488 goto out;
489 }
490
491 /*
492 * Keep track of amount of locked VM.
493 */
494 nr_pages = (end - start) >> PAGE_SHIFT;
495 if (!vma_flags_test(new_vma_flags, VMA_LOCKED_BIT))
496 nr_pages = -nr_pages;
497 else if (vma_flags_test(&old_vma_flags, VMA_LOCKED_BIT))
498 nr_pages = 0;
499 mm->locked_vm += nr_pages;
500
501 /*
502 * vm_flags is protected by the mmap_lock held in write mode.
503 * It's okay if try_to_unmap_one unmaps a page just after we
504 * set VMA_LOCKED_BIT, populate_vma_page_range will bring it back.
505 */
506 if (vma_flags_test(new_vma_flags, VMA_LOCKED_BIT) &&
507 vma_flags_test(&old_vma_flags, VMA_LOCKED_BIT)) {
508 /* No work to do, and mlocking twice would be wrong */
509 vma_start_write(vma);
510 vma->flags = *new_vma_flags;
511 } else {
512 mlock_vma_pages_range(vma, start, end, new_vma_flags);
513 }
514out:
515 *prev = vma;
516 return ret;
517}
518
519static int apply_vma_lock_flags(unsigned long start, size_t len,
520 const vma_flags_t *flags)
521{
522 unsigned long nstart, end, tmp;
523 struct vm_area_struct *vma, *prev;
524 VMA_ITERATOR(vmi, current->mm, start);
525
526 VM_BUG_ON(offset_in_page(start));
527 VM_BUG_ON(len != PAGE_ALIGN(len));
528 end = start + len;
529 if (end < start)
530 return -EINVAL;
531 if (end == start)
532 return 0;
533 vma = vma_iter_load(&vmi);
534 if (!vma)
535 return -ENOMEM;
536
537 prev = vma_prev(&vmi);
538 if (start > vma->vm_start)
539 prev = vma;
540
541 nstart = start;
542 tmp = vma->vm_start;
543 for_each_vma_range(vmi, vma, end) {
544 int error;
545 vma_flags_t newflags;
546
547 if (vma->vm_start != tmp)
548 return -ENOMEM;
549
550 newflags = vma->flags;
551 vma_flags_clear_mask(&newflags, VMA_LOCKED_MASK);
552 vma_flags_set_mask(&newflags, *flags);
553
554 /* Here we know that vma->vm_start <= nstart < vma->vm_end. */
555 tmp = vma->vm_end;
556 if (tmp > end)
557 tmp = end;
558 error = mlock_fixup(&vmi, vma, &prev, nstart, tmp, &newflags);
559 if (error)
560 return error;
561 tmp = vma_iter_end(&vmi);
562 nstart = tmp;
563 }
564
565 if (tmp < end)
566 return -ENOMEM;
567
568 return 0;
569}
570
571/*
572 * Go through vma areas and sum size of mlocked
573 * vma pages, as return value.
574 * Note deferred memory locking case(mlock2(,,MLOCK_ONFAULT)
575 * is also counted.
576 * Return value: previously mlocked page counts
577 */
578static unsigned long count_mm_mlocked_page_nr(struct mm_struct *mm,
579 unsigned long start, size_t len)
580{
581 struct vm_area_struct *vma;
582 unsigned long count = 0;
583 unsigned long end;
584 VMA_ITERATOR(vmi, mm, start);
585
586 /* Don't overflow past ULONG_MAX */
587 if (unlikely(ULONG_MAX - len < start))
588 end = ULONG_MAX;
589 else
590 end = start + len;
591
592 for_each_vma_range(vmi, vma, end) {
593 if (vma_test(vma, VMA_LOCKED_BIT)) {
594 if (start > vma->vm_start)
595 count -= (start - vma->vm_start);
596 if (end < vma->vm_end) {
597 count += end - vma->vm_start;
598 break;
599 }
600 count += vma->vm_end - vma->vm_start;
601 }
602 }
603
604 return count >> PAGE_SHIFT;
605}
606
607/*
608 * convert get_user_pages() return value to posix mlock() error
609 */
610static int __mlock_posix_error_return(long retval)
611{
612 if (retval == -EFAULT)
613 retval = -ENOMEM;
614 else if (retval == -ENOMEM)
615 retval = -EAGAIN;
616 return retval;
617}
618
619static __must_check int do_mlock(unsigned long start, size_t len,
620 vma_flags_t *flags)
621{
622 unsigned long locked;
623 unsigned long lock_limit;
624 int error = -ENOMEM;
625
626 start = untagged_addr(start);
627
628 if (!can_do_mlock())
629 return -EPERM;
630
631 len = PAGE_ALIGN(len + (offset_in_page(start)));
632 start &= PAGE_MASK;
633
634 lock_limit = rlimit(RLIMIT_MEMLOCK);
635 lock_limit >>= PAGE_SHIFT;
636 locked = len >> PAGE_SHIFT;
637
638 if (mmap_write_lock_killable(current->mm))
639 return -EINTR;
640
641 locked += current->mm->locked_vm;
642 if ((locked > lock_limit) && (!capable(CAP_IPC_LOCK))) {
643 /*
644 * It is possible that the regions requested intersect with
645 * previously mlocked areas, that part area in "mm->locked_vm"
646 * should not be counted to new mlock increment count. So check
647 * and adjust locked count if necessary.
648 */
649 locked -= count_mm_mlocked_page_nr(current->mm,
650 start, len);
651 }
652
653 /* check against resource limits */
654 if ((locked <= lock_limit) || capable(CAP_IPC_LOCK))
655 error = apply_vma_lock_flags(start, len, flags);
656
657 mmap_write_unlock(current->mm);
658 if (error)
659 return error;
660
661 error = __mm_populate(start, len, 0);
662 if (error)
663 return __mlock_posix_error_return(error);
664 return 0;
665}
666
667SYSCALL_DEFINE2(mlock, unsigned long, start, size_t, len)
668{
669 vma_flags_t flags = mk_vma_flags(VMA_LOCKED_BIT);
670
671 return do_mlock(start, len, &flags);
672}
673
674SYSCALL_DEFINE3(mlock2, unsigned long, start, size_t, len, int, flags)
675{
676 vma_flags_t vma_flags = mk_vma_flags(VMA_LOCKED_BIT);
677
678 if (flags & ~MLOCK_ONFAULT)
679 return -EINVAL;
680
681 if (flags & MLOCK_ONFAULT)
682 vma_flags_set(&vma_flags, VMA_LOCKONFAULT_BIT);
683
684 return do_mlock(start, len, &vma_flags);
685}
686
687SYSCALL_DEFINE2(munlock, unsigned long, start, size_t, len)
688{
689 vma_flags_t flags = EMPTY_VMA_FLAGS;
690 int ret;
691
692 start = untagged_addr(start);
693
694 len = PAGE_ALIGN(len + (offset_in_page(start)));
695 start &= PAGE_MASK;
696
697 if (mmap_write_lock_killable(current->mm))
698 return -EINTR;
699 ret = apply_vma_lock_flags(start, len, &flags);
700 mmap_write_unlock(current->mm);
701
702 return ret;
703}
704
705/*
706 * Take the MCL_* flags passed into mlockall (or 0 if called from munlockall)
707 * and translate into the appropriate modifications to mm->def_vma_flags and/or
708 * the flags for all current VMAs.
709 *
710 * There are a couple of subtleties with this. If mlockall() is called multiple
711 * times with different flags, the values do not necessarily stack. If mlockall
712 * is called once including the MCL_FUTURE flag and then a second time without
713 * it, VMA_LOCKED_BIT and VMA_LOCKONFAULT_BIT will be cleared from
714 * mm->def_vma_flags.
715 */
716static int apply_mlockall_flags(int flags)
717{
718 VMA_ITERATOR(vmi, current->mm, 0);
719 struct mm_struct *mm = current->mm;
720 struct vm_area_struct *vma, *prev = NULL;
721 vma_flags_t to_add = EMPTY_VMA_FLAGS;
722
723 vma_flags_clear_mask(&mm->def_vma_flags, VMA_LOCKED_MASK);
724 if (flags & MCL_FUTURE) {
725 vma_flags_set(&mm->def_vma_flags, VMA_LOCKED_BIT);
726
727 if (flags & MCL_ONFAULT)
728 vma_flags_set(&mm->def_vma_flags, VMA_LOCKONFAULT_BIT);
729
730 if (!(flags & MCL_CURRENT))
731 goto out;
732 }
733
734 if (flags & MCL_CURRENT) {
735 vma_flags_set(&to_add, VMA_LOCKED_BIT);
736 if (flags & MCL_ONFAULT)
737 vma_flags_set(&to_add, VMA_LOCKONFAULT_BIT);
738 }
739
740 for_each_vma(vmi, vma) {
741 int error;
742 vma_flags_t newflags = vma->flags;
743
744 vma_flags_clear_mask(&newflags, VMA_LOCKED_MASK);
745 vma_flags_set_mask(&newflags, to_add);
746
747 error = mlock_fixup(&vmi, vma, &prev, vma->vm_start, vma->vm_end,
748 &newflags);
749 /* Ignore errors, but prev needs fixing up. */
750 if (error)
751 prev = vma;
752 cond_resched();
753 }
754out:
755 return 0;
756}
757
758SYSCALL_DEFINE1(mlockall, int, flags)
759{
760 unsigned long lock_limit;
761 int ret;
762
763 if (!flags || (flags & ~(MCL_CURRENT | MCL_FUTURE | MCL_ONFAULT)) ||
764 flags == MCL_ONFAULT)
765 return -EINVAL;
766
767 if (!can_do_mlock())
768 return -EPERM;
769
770 lock_limit = rlimit(RLIMIT_MEMLOCK);
771 lock_limit >>= PAGE_SHIFT;
772
773 if (mmap_write_lock_killable(current->mm))
774 return -EINTR;
775
776 ret = -ENOMEM;
777 if (!(flags & MCL_CURRENT) || (current->mm->total_vm <= lock_limit) ||
778 capable(CAP_IPC_LOCK))
779 ret = apply_mlockall_flags(flags);
780 mmap_write_unlock(current->mm);
781 if (!ret && (flags & MCL_CURRENT))
782 mm_populate(0, TASK_SIZE);
783
784 return ret;
785}
786
787SYSCALL_DEFINE0(munlockall)
788{
789 int ret;
790
791 if (mmap_write_lock_killable(current->mm))
792 return -EINTR;
793 ret = apply_mlockall_flags(0);
794 mmap_write_unlock(current->mm);
795 return ret;
796}
797
798/*
799 * Objects with different lifetime than processes (SHM_LOCK and SHM_HUGETLB
800 * shm segments) get accounted against the user_struct instead.
801 */
802static DEFINE_SPINLOCK(shmlock_user_lock);
803
804int user_shm_lock(size_t size, struct ucounts *ucounts)
805{
806 unsigned long lock_limit, locked;
807 long memlock;
808 int allowed = 0;
809
810 locked = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
811 lock_limit = rlimit(RLIMIT_MEMLOCK);
812 if (lock_limit != RLIM_INFINITY)
813 lock_limit >>= PAGE_SHIFT;
814 spin_lock(&shmlock_user_lock);
815 memlock = inc_rlimit_ucounts(ucounts, UCOUNT_RLIMIT_MEMLOCK, locked);
816
817 if ((memlock == LONG_MAX || memlock > lock_limit) && !capable(CAP_IPC_LOCK)) {
818 dec_rlimit_ucounts(ucounts, UCOUNT_RLIMIT_MEMLOCK, locked);
819 goto out;
820 }
821 if (!get_ucounts(ucounts)) {
822 dec_rlimit_ucounts(ucounts, UCOUNT_RLIMIT_MEMLOCK, locked);
823 allowed = 0;
824 goto out;
825 }
826 allowed = 1;
827out:
828 spin_unlock(&shmlock_user_lock);
829 return allowed;
830}
831
832void user_shm_unlock(size_t size, struct ucounts *ucounts)
833{
834 spin_lock(&shmlock_user_lock);
835 dec_rlimit_ucounts(ucounts, UCOUNT_RLIMIT_MEMLOCK, (size + PAGE_SIZE - 1) >> PAGE_SHIFT);
836 spin_unlock(&shmlock_user_lock);
837 put_ucounts(ucounts);
838}